my learning rate was off in the first model. It was set to -1 and somehow it still worked, should of been .0004. So go figure. I did it again and looks the same. This version does not require a trigger word. Going to start making model with no trigger word so you can change the lora without changing the prompt.